Developer FAQ
Questions about privacy, pricing and the company are answered in the Loquela Remote FAQ.
Does Loquela Remote work with the OpenAI SDK?
Yes. Set the base URL to https://api.getloquela.com/v1 and use a Loquela Remote API key (starting lqr_). Model names such as whisper-1 and gpt-4o-transcribe are accepted, so the rest of your code can stay as it is. Switch from OpenAI
Which audio and video formats can I send?
WAV, MP3, M4A, AAC, OGG/Opus, WebM, FLAC, CAF and AIFF, and video files (MP4, MOV, 3GP, MKV, WebM). The type is read from the file itself, not its name. Transcribe
How long or large can a file be?
Up to 100 MB and 3 hours of audio per request. About 3 hours of speech fits in 100 MB as a 64 kbps MP3 or Opus file. Limits
Which languages does it transcribe?
English plus Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish and Ukrainian. English works on every account; the others are switched on per account from Languages on your account page. One language per request: send its code, or auto. unsupported_language
Can it tell speakers apart (diarization)?
Yes, for up to 8 speakers, labelled A to H in order of first appearance, at no extra cost. Turn it on with diarize=true or response_format=diarized_json. Speaker labels
Does it return word timestamps and subtitles?
Yes. verbose_json with timestamp_granularities[]=word gives each word with its start and end time, and srt and vtt return subtitles. Response formats
Are numbers written as digits?
Yes, by default. Spoken numbers come back as digits, for example 8.3% and $2.5 billion, while one to nine on their own, ordinals, clock times and phone numbers stay as words. Send written_numbers=false to keep the model's own wording. Numbers
Is my audio stored or used for training?
No. Audio is held in memory while it is transcribed, then erased with the text. Neither is stored, so neither is used for training. For billing we keep one line per request: the time, the audio's length, the minutes charged and whether it worked. How your audio is handled
Can I encrypt audio and transcripts with PGP?
Yes, on Pro and Business plans and Enterprise, at no extra cost. Encrypt the recording to your account's own PGP key before sending it, and send encrypt_to to get the transcript back signed by us and encrypted to your key, so only encrypted data passes between you and our computer. Our computer still decrypts the audio in memory to transcribe it. PGP encryption
How fast is it, and how accurate?
Over 400× faster than real time: an hour of audio takes under 9 seconds on our server, plus the time to upload it (our own test, 27 October 2026). Our current English model makes fewer mistakes than our previous one on every English test set we run. Speed and accuracy
Is there a realtime or streaming API?
No. Send a recording once it is finished; there is no live text while someone speaks. What differs from OpenAI
Can I use it from Claude, ChatGPT or Cursor?
Yes. The MCP server at https://api.getloquela.com/mcp gives AI assistants tools to transcribe audio from a link, an upload link or a short inline clip. MCP server
Can I transcribe without the API?
Yes. Sign in and open https://api.getloquela.com/transcribe: drop in files or record from your microphone, and get text or SRT and VTT subtitles back, from the same minutes as the API. Your audio is handled the same way. The difference is your browser: extensions that can read web pages can see the files and transcripts there, so the API is the more private route. With PGP on your plan, Expert mode encrypts audio in the browser and can return the transcript encrypted to your key. In your browser
How is usage billed?
By the second of audio, from your plan or prepaid minutes. Failed requests are free, and a partly failed file is charged only for what was transcribed. How minutes are counted
How do I check how many minutes I have left?
Call GET /v1/usage with your API key, or use the get_usage tool over MCP. The account page shows the same numbers. Usage
What happens when I run out of minutes or send too many requests?
The request is refused with status 429 and a code that says why, such as insufficient_quota or rate_limit_exceeded. When waiting helps, the Retry-After header says how long. Errors
Is there an Enterprise option for high volume?
Yes. For sustained, high-volume use, Enterprise rates start at $0.0008 a minute, with more requests at a time and per minute than Business, invoices and annual or committed-volume agreements, and help with security reviews. Email info@getloquela.com. Enterprise
Who can use Loquela Remote?
Customers in the United States. Loquela Remote is in open beta and close to release, and Founder program members get early access to new models, free credits and more. Overview